Google adds new TPU 8tx and TPU 8i Max for AI workloads
Both systems reuse existing TPU silicon, adding confidential computing or more memory bandwidth without a new chip generation.
TPU 8tx protects sensitive data and model weights during processing; Google claims up to twice TPU 8t’s performance with a near-zero security penalty. TPU 8i Max packages two TPU 8i chips together, doubling compute, SRAM and memory bandwidth for large-context agents, though Google has not disclosed its total high-bandwidth memory capacity. Pricing, supported regions and general availability dates remain unspecified.
TestingCatalog's note
TPU 8tx protects sensitive data and model weights during processing; Google claims up to twice TPU 8t’s performance with a near-zero security penalty. TPU 8i Max packages two TPU 8i chips together, doubling compute, SRAM and memory bandwidth for large-context agents, though Google has not disclosed its total high-bandwidth memory capacity. Pricing, supported regions and general availability dates remain unspecified.
TestingCatalog's note
score 5